Joint analysis of individual participants data from 17...

ORIGINAL ARTICLE

Joint analysis of individual participants’ data from 17 studies on theassociation of the IL6 variant -174G�C with circulating glucoselevels, interleukin-6 levels, and body mass index

CORNELIA HUTH1,2, THOMAS ILLIG1, CHRISTIAN HERDER3,

CHRISTIAN GIEGER1,2, HARALD GRALLERT1, CAREN VOLLMERT1,4,

WOLFGANG RATHMANN3, YASMIN H. HAMID5, OLUF PEDERSEN5,

TORBEN HANSEN5, BARBARA THORAND1, CHRISTA MEISINGER1,

ANGELA DÖRING1, NORMAN KLOPP1, HENNING GOHLKE1, WOLFGANG LIEB6,

CHRISTIAN HENGSTENBERG7, VALERIYA LYSSENKO8, LEIF GROOP8,

HELEN IRELAND9, JEFFREY W. STEPHENS10, INGRID WERNSTEDT ASTERHOLM11,

JOHN-OLOV JANSSON11, HEINER BOEING12, MATTHIAS MÖHLIG13,

HEATHER M. STRINGHAM14, MICHAEL BOEHNKE14, JAAKKO TUOMILEHTO15�17,

JOSE-MANUEL FERNANDEZ-REAL18, ABEL LOPEZ-BERMEJO18, LUIS GALLART19,

JOAN VENDRELL19, STEVE E. HUMPHRIES9, FLORIAN KRONENBERG20,

H.-ERICH WICHMANN1,2 & IRIS M. HEID1,2

1Institute of Epidemiology, Helmholtz Zentrum München, Neuherberg, Germany, 2Institute of Biometry and Epidemiology,

University of Munich, Germany, 3German Diabetes Center, Leibniz Institute at Heinrich Heine University Düsseldorf,

Germany, 4Sequenom GmbH, Hamburg, Germany, 5Steno Diabetes Center, Copenhagen, Denmark, 6Clinic and Policlinic

for Internal Medicine II and Institute of Human Genetics, University of Lübeck, Germany, 7Clinic and Policlinic for Internal

Medicine II, University of Regensburg, Germany, 8Department of Clinical Sciences, University Hospital Malmö, Sweden,9Centre for Cardiovascular Genetics, Royal Free and University College Medical School, London, UK, 10Medical School,

University of Wales, Swansea, UK, 11Institute of Neuroscience and Physiology, Sahlgrenska Academy at Gothenborg

University, Sweden, 12Department of Epidemiology, German Institute of Human Nutrition, Potsdam-Rehbruecke, Germany,13Department of Clinical Nutrition, German Institute of Human Nutrition, Potsdam-Rehbruecke, Germany, 14Department of

Biostatistics, University of Michigan, USA, 15Diabetes Unit, National Public Health Institute, Helsinki, Finland,16Department of Public Health, University of Helsinki, Finland, 17South Ostrobothnia Central Hospital, Seinäjoki, Finland,18Section of Diabetes, Endocrinology and Nutrition, University Hospital of Girona Dr Josep Trueta and ‘CIBER

Fisiopatologı́a de la Obesidad y Nutrición’, Girona, Spain, 19Research Unit, University Hospital Joan XXIII, Pere Virgili

Institute and CIBERDEM, Tarragona, Spain, and 20Division of Genetic Epidemiology, Innsbruck Medical University,

Austria

AbstractBackground. Several studies have investigated associations between the -174G�C single nucleotide polymorphism(rs1800795) of the IL6 gene and phenotypes related to type 2 diabetes mellitus (T2DM) but presented inconsistent results.Aims. This joint analysis aimed to clarify whether IL6 -174G�C was associated with glucose and circulating interleukin-6concentrations as well as body mass index (BMI).

Correspondence: Iris M. Heid, Institute of Epidemiology, Helmholtz Zentrum München*German Research Center for Environmental Health, IngolstädterLandstrasse 1, D-85764 Neuherberg, Germany. Fax: �49 89 3187 3380. E-mail: [email protected]

(Received 19 February 2008; revised 27 May 2008; accepted 9 July 2008)

Annals of Medicine. 2009; 41: 128�138

ISSN 0785-3890 print/ISSN 1365-2060 online # 2009 Informa UK Ltd. (Informa Healthcare, Taylor & Francis AS)DOI: 10.1080/07853890802337037

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Methods. Individual-level data from all studies of the IL6-T2DM consortium on Caucasian subjects with available BMI werecollected. As study-specific estimates did not show heterogeneity (P�0.1), they were combined by using the inverse-variance fixed-effect model.Results. The main analysis included 9440, 7398, 24,117, or 5659 non-diabetic and manifest T2DM subjects for fastingglucose, 2-hour glucose, BMI, or circulating interleukin-6 levels, respectively. IL6 -174 C-allele carriers had significantlylower fasting glucose (�0.091 mmol/L, P�0.014). There was no evidence for association between IL6 -174G�C andBMI or interleukin-6 levels, except in some subgroups.Conclusions. Our data suggest that C-allele carriers of the IL6 -174G�C polymorphism have lower fasting glucose levels onaverage, which substantiates previous findings of decreased T2DM risk of these subjects.

Key words: Blood glucose, body mass index, diabetes mellitus type 2, genes, inflammation mediators, interleukin-6,intermediate phenotype, meta-analysis, molecular epidemiology, single nucleotide polymorphism

Introduction

Type 2 diabetes mellitus (T2DM) is a public health

problem of pandemic proportions. Conservative

estimates indicate that there are 171 million people

in the world with symptomatic or asymptomatic

diabetes mellitus and that this global prevalence will

double by 2030 to 366 million people (1). The

spread of the disease is also alarming in Europe, even

though Caucasians have a low to moderate preva-

lence of T2DM compared with most other ethnic

groups worldwide (2).

Type 2 diabetes mellitus is a multifactorial

disease. While a lot is known about environmental

risk factors for T2DM, identification of the genetic

etiology of T2DM has proven to be challenging (3).

An interesting candidate variant for T2DM is the

-174G�C polymorphism (rs1800795) of the IL6gene that codes for the cytokine interleukin-6 (IL-6).

Interleukin-6 exerts pleiotropic biological functions

in the regulation of the acute-phase reaction and

immune responses, and is associated with T2DM

and related diseases (4�7). The promoter poly-morphism IL6 -174G�C, which has been shownto affect IL6 promoter activity (8,9), was reported to

be associated with circulating IL-6 levels (8,10�14),T2DM (15), insulin resistance (16), obesity (17�20), coronary heart disease (CHD) (21), cholesterol

levels (6), and metabolic syndrome (22). However,

association estimates pointed in inconclusive direc-

tions, while other studies of these traits did not find

any association.

At the start of our project, meta-analyses on

association between IL6 -174G�C and two differentoutcomes had been reported: 1) Sie et al. did not find

statistically significant evidence for an association

with CHD in 19,798 individuals (23). Having large

heterogeneity between association estimates, Qi et al.

did not detect a statistically significant association

with T2DM in 17,452 individuals of mainly pub-

lished studies (24). We established an international

IL6-T2DM consortium on individual-level published

and unpublished data and showed an almost 9%

reduced T2DM risk for C-allele carriers in 20,976

individuals (25). However, it was not clear yet

whether the IL6 -174G�C polymorphism impactsIL-6 levels which together with the T2DM associa-

tion would support a causal role of IL-6 levels in

T2DM. Moreover, studies on genetic association

with obesity pointed in a direction that was opposite

to the T2DM association (17,18). Finally, analyzing

the dichotomous trait T2DM is a substantial restric-

tion of information, which can be overcome by

analyzing the quantitative trait circulating plasma or

serum glucose in the fasting state or after an oral

Key messages

. Several studies have investigated associa-tions between the -174G�C polymorph-ism (rs1800795) of the IL6 gene but

presented inconsistent results. For all ana-

lyzed phenotypes, our joint analysis repre-

sents the largest study on individual-level

data conducted to date to address the role

of IL6 -174G�C.. We were able to reveal lower fasting glucose

levels in carriers of the IL6 -174 C-allele.

There was no evidence for association

between IL6 -174G�C and BMI or circu-lating interleukin-6 levels, except in some

subgroups.

Abbreviations

BMI body mass index

CHD coronary heart disease

HWE Hardy-Weinberg equilibrium

IL-6 interleukin-6

IL6 interleukin-6 gene

OR odds ratio

T2DM type 2 diabetes mellitus

2-h 2-hour

95% CI 95% confidence interval

IL6*174G�C and quantitative phenotypes 129

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

glucose load. This is of special importance because

the T2DM finding was borderline statistically sig-

nificant (P�0.037) (25), and a consistent glucoseassociation would greatly underscore this finding.

The objective of the present study was to clarify

the epidemiological evidence for the association of

IL6 -174G�C polymorphism with circulating glu-cose levels, body mass index (BMI), and IL-6 levels as

intermediate phenotypes on the way to T2DM. We

used by far the largest individual-level data set on this

polymorphism in relation to quantitative phenotypes

to date. It is important to note that earlier studies

detected associations between -174G�C and IL-6levels or obesity mostly in the presence of an

inflammatory stimulus (10,11) or in subjects having

a chronic subclinical (19), or an acute (12,14)

inflammatory state. The inclusion of T2DM studies

in our joint analysis was essential to investigate

whether the impact of IL6 -174G�C indeed differedby the subclinical inflammatory state observed in type

2 diabetic individuals, as opposed to healthy subjects.

Subjects and methods

All studies of the IL6-T2DM consortium on Cau-

casian subjects with available BMI were included in

this joint analysis of quantitative traits. Information

on the inclusion criteria, the search strategy and

recruitment of the studies, data cleaning, and

genotyping methods is provided in the online

appendix.

Definition of analyzed samples and data collection

Because the majority of included studies were cross-

sectional population-based or T2DM case-control

studies, our main analysis focused on a combined

sample of non-diabetic and prevalent T2DM sub-

jects. We included the baseline examination of the

two longitudinal cohort studies excluding subjects

that developed T2DM during follow-up for this

main analysis. All analyses were confined to Cauca-

sian adults who were at least 18 years old with data

on the genotype of the IL6 -174G�C polymorph-ism, age, sex, BMI, and T2DM status. When

known, type 1 diabetic individuals were excluded.

In family studies, only the sibling generation was

used. Data on circulating fasting and 2-hour plasma

or serum glucose (measured two hours after con-

sumption of 75 g glucose) and plasma or serum IL-6

levels were collected where available.

All participating studies have been conducted

according to the principles expressed in the Declara-

tion of Helsinki. Individual studies had either written

informed consent for all subjects for genetic analyses

or approval from their institutional review commit-

tee for genetic analyses. Further details on design of

individual studies and use in this joint analysis are

presented in the online appendix Table A-II.

Statistical analyses

Study-specific b-coefficients for the association be-tween IL6 -174G�C and the quantitative traitswere estimated by linear regression, using SAS

PROC GLM for studies with unrelated individuals

and SAS PROC MIXED for family studies. All

analyses were adjusted for age and sex. Primary

analyses were performed including all subjects (full

data analyses) additionally adjusting for T2DM

status when analyzing BMI or IL-6 levels. Secondary

analyses were conducted separately for the non-

diabetic and the prevalent T2DM subjects (T2DM

status-specific analyses), additionally adjusting for

BMI when analyzing fasting, 2-hour glucose, or IL-6

levels. If a study featured less than 50 participants

for a specific analysis, this study was not included in

the joint analysis. Studies with fasting glucose only

on either T2DM cases or controls (EDSC, EPIC-

POTSDAM_nCC-T2DM, and MONICA-S3) were

excluded from the full data glucose analysis to assure

a full range of glucose levels.

IL6 -174G�C genotypes were analyzed model-free, comparing CC- or GC-genotype with the wild-

type GG, and additionally by applying a dominant

genetic model for the C-allele, which was the model

reported previously for the T2DM association (25).

Heterogeneity between study-specific b-coefficientswas tested by the chi-square-based Q-statistic, and

its impact was quantified by I2 (26). As the hetero-

geneity between study-specific b-coefficients wasnon-significant in all analyses (P�0.10), the b-coefficients were combined using the inverse-var-

iance fixed-effect model (27). As recommended,

summary association estimates of all studies with

the genotype frequencies of non-diabetic subjects

being in Hardy-Weinberg equilibrium (HWE) were

reported as main results (28).

A conservative Bonferroni-corrected significance

level of 0.017 (�0.05/3) was applied to account forthe testing of the three phenotypes (glucose, BMI,

and IL-6) in the primary (full data) analyses. Note

that the phenotypes fasting and 2-hour postprandial

glucose levels were highly correlated and thus not

counted as two independent phenotypes here. In the

secondary analyses, the significance level was further

corrected for the two investigated subgroups yielding

a significance level of 0.008 (0.05/6). More detailed

information on statistical procedures is given in the

online appendix.

130 C. Huth et al.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Analysis in incident T2DM patients

In an additional exploratory analysis, we combined

the linear regression estimates of baseline data from

the incident T2DM cases of the two cohort studies

to obtain estimates of the IL6 -174G�C associationwith BMI and IL-6 levels (glucose not available)

among future T2DM subjects before they became

cases (‘prediabetic subjects’).

Results

Study recruitment

Seventeen studies with 25,635 participants met the

study and subject inclusion criteria and were in-

cluded in the joint analyses (see Table I for an

overview). Online appendix Table A-III shows for

which outcome the respective study qualified for

analysis; online appendix Table A-IV presents the

number of participants per study included in the

analyses of the respective quantitative trait. All 17

studies with 25,635 subjects were analyzed for the

outcome BMI. Association analyses with BMI as

main outcome had been unpublished in 12 studies at

the time of study recruitment (called ‘unpublished

for BMI’ in the following). Eight studies with 10,725

participants were analyzed for fasting glucose, seven

of them unpublished for this trait; seven studies with

8399 participants were analyzed for 2-hour glucose

(all unpublished). For the outcome circulating IL-6

levels, data from seven studies with 5659 partici-

pants were analyzed, six of them unpublished for IL-

6 levels.

Study-specific statistics

Detailed characteristics of included studies and

participants are summarized in online appendix

Table A-V. All studies had been conducted in

European populations. Genotype frequencies of

non-diabetic subjects were in HWE for all studies,

except for the BOTNIA and the TGN study, which

were thus excluded from the main analyses. This left

9440, 7398, 24,117, and 5659 subjects for the

analysis of fasting glucose, 2-hour glucose, BMI,

and circulating IL-6, respectively (see online appen-

dix Table A-IV). The IL6 -174 C-allele frequency of

non-diabetic subjects with genotype frequencies in

HWE ranged from 41.1% (95% CI�38.3�43.8) inthe KORA-MIFAM study to 55.0% (95% CI�51.4�58.6) in the FUSION 1 study.

IL6 -174G�C and circulating glucose

The outcome fasting glucose was investigated

as full data main analysis for the seven studies

where the genotype frequencies of non-diabetic

Table I. Characteristics of included studies.

Studya Full study name Countryb n T2DM/non-diabetic subjectsc

BOTNIA Botnia Study SF 731/557

CAPPP Captopril Prevention Project S 42/424

DANISH Danish Study DK 1212/4399

EDSC Ealing Diabetes Study of Coagulation UK 299/0

EPIC-POTSDAM European Prospective Investigation into Cancer and Nutrition

Potsdam (EPIC-Potsdam)

D 0/348

FUSION 1 The Finland-United States Investigation of NIDDM Genetics,

1st sampling wave

SF 508/367

FUSION 2 The Finland-United States Investigation of NIDDM Genetics,

2nd sampling wave

SF 437/201

GIRONA Girona Genetics of Diabetes Study E 42/123

KORA-MIFAM KORA MI Family Study D 95/881

KORA-S4 KORA Survey S4 D 225/1190

KORA-T2DMFAM KORA T2DM Family Study D 776/513

MONICA/KORA-BASE MONICA/KORA Case Cohort Study S123

(MONICA/KORA-S123)

D 101/1744

MONICA-S3 MONICA/KORA Survey S3 D 151/3551

NPHS II Second Northwick Park Heart Study UK 0/2652

RMIFAM Regensburg Ml Family Study D 662/2614

TGN Tarraco Study E 166/64

UDACS University College Diabetes and Cardiovascular Study UK 560/0

aAbbreviated study name used in the present publication.bCountry of recruitment: D�Germany, DK�Denmark, E�Spain, SF�Finland, S�Sweden, UK�United Kingdom.cNumber of type 2 diabetic/non-diabetic subjects included in analyses of the outcome BMI.

MI�myocardial infarction; NIDDM�non-insulin-dependent diabetes mellitus.


Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

subjects were in HWE, including 9440 subjects.

Figure 1 shows the study-specific and the summary

b-coefficients, using a dominant model for theC-allele. Most studies exhibited a reduction of

fasting glucose levels, though not significant in

each study separately. The summary estimate

provided a statistically significant decrease of

�0.091 mmol/L (95% CI�(�0.163)�(�0.018),P�0.014), corresponding to a decrease of about1.5% for subjects with GC- or CC-genotypes

compared to GG. T2DM status-specific analyses

showed a more pronounced reduction of �0.317mmol/L (95% CI�(�0.584)�(�0.049), P�0.020)among T2DM subjects.

Table II provides model-free summary estimates,

indicating that the dominant genetic model was

consistent with the data.

The full data main analysis of 2-hour glucose

included six studies with 7398 subjects. There was

no statistically significant association between IL6

-174G�C and 2-hour glucose (C-allele dominantmodel for full data analysis: b��0.075 mmol/L,95% CI�(�0.201)�(0.051), P�0.243) (Table II).

IL6 -174G�C and body mass index

The full data main analysis of the outcome BMI

included 15 studies with 24,117 subjects. Figure 2

Study Association Estimate ß [95% CI] Weight [%]

(A) Full Data Analysis GIRONA 1.33 -0.58 [-1.20, 0.05] FUSION 2 1.40 -0.53 [-1.14, 0.08] FUSION 1 1.57 -0.50 [-1.07, 0.08] CAPPP 2.55 -0.03 [-0.48, 0.42] KORA-T2DMFAM 11.78 0.11 [-0.10, 0.32] DANISH 38.92 -0.08 [-0.20, 0.03] KORA-S4 42.43 -0.11 [-0.22, 0.00]

Summary Association Estimate β β 100.00 -0.09 [-0.16, -0.02]Test for heterogeneity: Chi² = 9.79, df = 6 (P = 0.13), I² = 38.7%Test for overall effect: Z = 2.45 (P = 0.01)

(B) Nondiabetic Individuals EPIC-POTSDAM 0.24 0.29 [-0.13, 0.71] GIRONA 0.78 -0.16 [-0.40, 0.07] CAPPP 0.93 -0.14 [-0.35, 0.08] FUSION 2 1.20 0.04 [-0.15, 0.23] MONICA-S3 2.14 -0.02 [-0.16, 0.12] FUSION 1 2.78 -0.01 [-0.13, 0.12] KORA-T2DMFAM 5.47 0.05 [-0.03, 0.14] KORA-S4 11.40 -0.01 [-0.07, 0.05] DANISH 75.06 0.01 [-0.01, 0.03]

Summary Association Estimate β β 100.00 0.01 [-0.01, 0.03] Test for heterogeneity: Chi² = 7.28, df = 8 (P = 0.51), I² = 0%Test for overall effect: Z = 0.67 (P = 0.50)

(C) Type 2 Diabetic Individuals EDSC 2.30 0.78 [-0.98, 2.55] FUSION 2 13.05 -0.55 [-1.29, 0.19] FUSION 1 14.09 -0.42 [-1.13, 0.30] KORA-T2DMFAM 17.03 0.31 [-0.34, 0.95] KORA-S4 17.08 -0.87 [-1.51, -0.22] DANISH 36.45 -0.30 [-0.74, 0.14]

Summary Association Estimate β β 100.00 -0.32 [-0.58, -0.05] Test for heterogeneity: Chi² = 8.26, df = 5 (P = 0.14), I² = 39.5%Test for overall effect: Z = 2.32 (P = 0.02)

-2 -1 0 1 2 CC+GC lower glucose CC+GC higher glucose

Association Estimate ß [95% CI]

(mmol/L)

Figure 1. Forest plot, illustrating the study-specific b-coefficients with 95% CIs for the association between IL6 -174G�C, dominantmodel for the C-allele, and fasting glucose in (A) the full data analysis (combined group of non-diabetic and prevalent type 2 diabetes

mellitus (T2DM) subjects), (B) only in non-diabetic subjects, and (C) only in prevalent T2DM subjects, adjusted for age, sex, and body

mass index (BMI). Additionally, the summary fixed-effect ß-coefficient is shown. I2 measures the impact of inconsistency across studies and

can range between 0% and 100%.

132 C. Huth et al.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

shows the study-specific and the summary b-coeffi-cients applying the dominant model for the C-allele.

There was no statistically significant evidence for an

association between IL6 -174G�C and BMI in thefull data (b�0.088, 95% CI�(�0.023)�(0.200),P�0.120) or the T2DM status-specific analyses(P�0.10) (Table II).

IL6 -174G�C and circulating interleukin-6 levels

There was no evidence in the full data joint analysis

including 5659 subjects of seven studies for an

association between IL6 -174G�C and circulatingIL-6 levels (Table II and Figure 3). Likewise, there

was no statistically significant IL-6 level association

among the 4621 non-diabetic subjects of six studies

or among the 966 prevalent T2DM subjects of four

studies.

Analysis of BMI and IL-6 levels in incident T2DM cases

before they became cases

Among the 641 incident T2DM subjects before they

became type 2 diabetic, there was no association

with BMI (P�0.10). Regarding circulating IL-6, theIL6 -174G�C showed no association in the domi-nant model, but higher levels for the homozygous

CC-genotype (bCCvsGG�0.266, 95% CI�0.085�0.448, P�0.004) compared to the GG-genotype.

Sensitivity analyses

Regarding fasting glucose, 2-hour glucose, or IL-6

levels, there was no evidence for publication bias: the

Egger’s regression test was non-significant (P�0.10), and excluding published studies would not

change the main findings (online appendix Table A-

VI). Regarding BMI, there was some evidence for

publication bias. In the five published studies with

10,704 subjects, the IL6 -174 C-allele was statisti-

cally significantly associated with higher BMI (b�0.190, 95% CI�0.023�0.358). In contrast,there was no association in the ten unpublished

studies with 13,413 subjects (b�0.007, 95% CI�(�0.142)�(0.157)). Moreover, the funnel plot (on-line appendix Figure A-1) and the Egger’s regression

test (P�0.06) showed some evidence for publica-tion bias of the summary BMI estimate when

including all 15 studies, but none when restricting

to the ten unpublished studies (P�0.27).Including studies with genotype HWE violation

(online appendix Table A-VII) or omitting adjust-

ment for BMI or T2DM status (data not shown)

would not change our main findings. Furthermore,

there were no large differences between associationTable

II.S

um

mary

resu

lts

of

ass

oci

ati

on

bet

wee

nIL

6-1

74G�

Can

dfa

stin

gglu

cose

,2-h

glu

cose

,b

od

ym

ass

ind

ex(B

MI)

,or

inte

rleu

kin

-6(I

L-6

)le

vels

.T

he

fixed

-eff

ect

resu

lts

are

pre

sen

ted

for

the

full

data

an

aly

sisa

an

dfo

ran

aly

ses

stra

tifi

edfo

rty

pe

2d

iabet

esst

atu

s(T

2D

M).

Fu

lld

ata

an

aly

sisa

Non

-dia

bet

icsu

bje

cts

Type

2d

iabet

icsu

bje

cts

nß

-coef

fici

ent

(P-v

alu

e)b

nß

-coef

fici

ent

(P-v

alu

e)b

nß

-coef

fici

ent

(P-v

alu

e)b

Ou

tcom

eS

tud

ies

Su

bje

cts

CC

vs

GG

GC

vs

GG

Dom

inan

tcS

tud

ies

Su

bje

cts

CC

vs

GG

GC

vs

GG

Dom

inan

tbS

tud

ies

Su

bje

cts

CC

vs

GG

GC

vs

GG

Dom

inan

tb

Fast

ing

glu

cose

(mm

ol/L

)

79440

�0.0

81

(0.1

08)

�0.0

99

(0.0

11)

�0.0

91

(0.0

14)

97420

0.0

07

(0.6

26)

0.0

07

(0.5

23)

0.0

07

(0.5

01)

62412

�0.3

22

(0.0

76)

�0.3

54

(0.0

13)

�0.3

17

(0.0

20)

2-h

ou

rglu

cose

(mm

ol/L

)

67398

�0.1

27

(0.1

27)

�0.0

53

(0.4

41)

�0.0

75

(0.2

43)

66731

�0.0

22

(0.5

92)

�0.0

03

(0.9

20)

�0.0

09

(0.7

75)

4618

�0.0

43

(0.9

27)

�0.1

14

(0.7

60)

�0.1

35

(0.7

03)

BM

I(k

g/m

2)

15

24117

0.0

97

(0.1

97)

0.0

86

(0.1

51)

0.0

88

(0.1

20)

13

19007

0.0

57

(0.4

78)

0.0

80

(0.2

13)

0.0

71

(0.2

42)

11

5026

0.2

16

(0.2

55)

0.1

54

(0.3

21)

0.1

76

(0.2

26)

Cir

cula

tin

gIL

-6d

75659

0.0

34

(0.2

92)

0.0

18

(0.4

67)

0.0

24

(0.3

04)

64621

0.0

01

(0.9

69)

0.0

22

(0.4

41)

0.0

18

(0.5

04)

4966

0.1

59

(0.0

44)

0.0

04

(0.9

38)

0.0

42

(0.4

15)

aA

naly

sis

ofn

on

-dia

bet

ican

dpre

vale

nt

T2D

Msu

bje

cts

toget

her

.T

he

sum

ofth

en

um

ber

ofn

on

-dia

bet

ican

dT

2D

Msu

bje

cts

inst

rati

fied

an

aly

ses

does

not

yie

ldth

en

um

ber

ofsu

bje

cts

inth

efu

ll

data

an

aly

sis,

bec

au

sest

rati

fied

an

aly

ses

wer

eper

form

edin

ind

ivid

ual

stu

die

son

lyif

ther

ew

ere

at

least

50

non

-dia

bet

icor

T2D

Msu

bje

cts

availab

le.

bS

um

mary

esti

mate

sfr

om

gen

eralize

dlin

ear

mod

els

ad

just

edfo

rage

an

dse

x(f

or

BM

Ian

dIL

-6le

vels

:ad

dit

ion

ally

ad

just

edfo

rT

2D

Mst

atu

s;fo

rglu

cose

an

dIL

-6le

vel

s:ad

dit

ion

ally

ad

just

edfo

r

BM

I).

Main

fin

din

gs

are

pri

nte

din

bold

.cD

om

inan

tm

od

elco

mpari

ng

CC

-an

dG

C-s

ubje

cts

vers

us

GG

.dIL

-6ß

-coef

fici

ents

wer

eco

mp

ute

dan

dare

dis

pla

yed

on

the

logari

thm

of

the

ori

gin

al

scale

(pg/m

L).


Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

GIRONA 0.57 -0.52 [-2.00, 0.96]EDSC 0.71 1.03 [-0.30, 2.35]UDACS 1.50 0.86 [-0.05, 1.77]EPIC-POTSDAM 1.76 0.16 [-0.69, 1.00]FUSION 2 2.09 0.19 [-0.58, 0.96]CAPPP 2.10 1.03 [0.26, 1.80]FUSION 1 2.59 0.36 [-0.33, 1.05]KORA-T2DMFAM 3.11 -0.16 [-0.79, 0.47]KORA-MIFAM 4.14 0.21 [-0.34, 0.76]KORA-S4 5.60 0.13 [-0.35, 0.60]MONICA/KORA-BASE 9.22 -0.14 [-0.50, 0.23]MONICA-S3 14.88 0.15 [-0.14, 0.44]NPHS II 15.93 0.22 [-0.06, 0.50]RMIFAM 16.55 -0.18 [-0.45, 0.10]DANISH 19.25 0.04 [-0.22, 0.29]

Summary Association Estimate β β 100.00 0.09 [-0.02, 0.20]Test for heterogeneity: Chi² = 18.90, df = 14 (P = 0.17), I² = 25.9%Test for overall effect: Z = 1.55 (P = 0.12)

(B) Nondiabetic individuals GIRONA 0.50 0.23 [-1.46, 1.91]FUSION 2 1.02 0.13 [-1.05, 1.30]FUSION 1 1.32 0.48 [-0.55, 1.52]KORA-T2DMFAM 1.78 0.32 [-0.57, 1.22]EPIC-POTSDAM 2.00 0.16 [-0.69, 1.00]CAPPP 2.19 0.98 [0.18, 1.79]KORA-MIFAM 4.44 0.30 [-0.27, 0.86]KORA-S4 5.55 0.15 [-0.35, 0.66]MONICA/KORA-BASE 10.17 -0.19 [-0.56, 0.19]RMIFAM 16.08 -0.21 [-0.51, 0.09]MONICA-S3 16.36 0.15 [-0.14, 0.45]NPHS II 18.13 0.22 [-0.06, 0.50]DANISH 20.44 -0.01 [-0.28, 0.25]

Summary Association Estimate β β 100.00 0.07 [-0.05, 0.19] Test for heterogeneity: Chi² = 13.76, df = 12 (P = 0.32), I² = 12.8%Test for overall effect: Z = 1.17 (P = 0.24)

(C) Type 2 Diabetic Individuals KORA-MIFAM 1.92 -0.87 [-2.92, 1.19]MONICA/KORA-BASE 2.31 1.05 [-0.82, 2.93]MONICA-S3 3.45 -0.13 [-1.67, 1.40]EDSC 4.62 1.03 [-0.30, 2.35]KORA-S4 5.23 0.01 [-1.24, 1.26]FUSION 1 9.45 0.30 [-0.62, 1.23]UDACS 9.81 0.86 [-0.05, 1.77]FUSION 2 10.59 0.22 [-0.66, 1.09]KORA-T2DMFAM 12.36 -0.34 [-1.15, 0.47]DANISH 18.40 0.32 [-0.34, 0.99]RMIFAM 21.86 -0.13 [-0.74, 0.48]

Summary Association Estimate β β 100.00 0.18 [-0.11, 0.46] Test for heterogeneity: Chi² = 8.62, df = 10 (P = 0.57), I² = 0%Test for overall effect: Z = 1.21 (P = 0.23)

-2 -1 0 1 2

CC+GC lower BMI CC+GC higher BMI

Study Association Estimate ß [95% CI] Weight [%] Association Estimate ß [95% CI]

(A) Full Data Analysis

(kg/m²)

Figure 2. Forest plot, illustrating the study-specific ß-coefficients with 95% CIs for the association between IL6 -174G�C, dominantmodel for the C-allele, and body mass index (BMI) of (A) the full data analysis, (B) only non-diabetic subjects, and (C) only prevalent type

2 diabetes mellitus (T2DM) subjects, adjusted for age, sex, and T2DM status (A). Please refer to Figure 1 legend for more details.

134 C. Huth et al.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

estimates of men and women in sex-stratified

analyses (data not shown).

Sensitivity analyses using ln-transformed fasting

glucose levels, ln-transformed BMI levels, or age- and

sex-preadjusted standard z-scores (all outcomes) to

obtain phenotype distributions which are closer to

normal yielded results which were very similar to the

main analyses. Likewise, exclusion of outliers (fasting

glucose B2.78 or �20.00 mmol/L, full data analysis:n�47; BMI B16.0 or �50.0 kg/m2, full dataanalysis: n�28; IL-6 levels B0.20 or �20.0 pg/mL,full data analysis: n�250, prevalent T2DM subjects:n�30) did not change the main results, except for astronger association between IL-6 levels (natural

logarithm) and the IL6 -174 CC- versus GG-geno-

type (b�0.226, 95% CI: 0.098�0.355, P�0.0006).Pooling the individual data rather than using a meta-

analysis approach yielded very similar results for all

analyses. Interestingly, a subgroup analysis with

exclusion of the two studies (DANISH and RMI-

FAM) with self-reported BMI data increased the

BMI association estimate (C-allele dominant model:

b�0.172 kg/m2, 95% CI: 0.033�0.312, P�0.015).

Discussion

Major findings

In this joint analysis of individual-level data from

9440 study participants, C-allele carriers of the IL6

-174G�C polymorphism had lower fasting glucoselevels (�0.091 mmol/L, P�0.014) independently ofBMI. No association was found with BMI in 24,117

subjects or with IL-6 levels in 5659 subjects. The

estimates were robust against exclusion of outliers,

various transformations of variables including use of

GIRONA 1.17 -0.05 [-0.65, 0.54]EPIC-POTSDAM 7.23 -0.12 [-0.36, 0.12]CAPPP 10.17 -0.15 [-0.35, 0.05]UDACS 11.66 0.19 [0.01, 0.38]KORA-S4 14.54 0.03 [-0.14, 0.20]MONICA/KORA-BASE 22.16 0.11 [-0.02, 0.25]KORA-MIFAM 33.06 0.02 [-0.09, 0.13]

100.00 0.03 [-0.03, 0.10] Test for heterogeneity: Chi² = 8.90, df = 6 (P = 0.18), I² = 32.6%Test for overall effect: Z = 1.05 (P = 0.29)

GIRONA 1.01 0.00 [-0.71, 0.71] EPIC-POTSDAM 8.87 -0.12 [-0.36, 0.12] CAPPP 11.04 -0.18 [-0.39, 0.04] KORA-S4 16.19 -0.03 [-0.20, 0.15] MONICA/KORA-BASE 25.00 0.12 [-0.02, 0.26] KORA-MIFAM 37.88 0.01 [-0.10, 0.13]

100.00 0.00 [-0.07, 0.07] Test for heterogeneity: Chi² = 6.49, df = 5 (P = 0.26), I² = 22.9%Test for overall effect: Z = 0.04 (P = 0.97)

KORA-S4 9.33 0.35 [-0.15, 0.86]KORA-MIFAM 10.29 0.03 [-0.45, 0.51]MONICA/KORA-BASE 12.52 -0.07 [-0.50, 0.37]UDACS 67.85 0.19 [0.01, 0.38]

100.00 0.16 [0.00, 0.31] Test for heterogeneity: Chi² = 2.00, df = 3 (P = 0.57), I² = 0%Test for overall effect: Z = 2.01 (P = 0.04)

-1 -0.5 0 0.5 1CC versus GG lower IL-6 CC versus GG higher IL-6

Study Association Estimate ß [95% CI] Weight [%] Association Estimate ß [95% CI]

(A) Full Data Analysis

Summary Association Estimate β β

(B) Nondiabetic Individuals


(C) Type 2 Diabetic Individuals


ln of (pg/mL)

Figure 3. Forest plot, illustrating the study-specific ß-coefficients with 95% CIs for the association between IL6 -174 CC versus GG and

natural logarithm of circulating interleukin-6 (IL-6) levels of (A) the full data analysis, (B) only non-diabetic subjects, and (C) only

prevalent type 2 diabetes mellitus (T2DM) subjects, adjusted for age, sex, body mass index (BMI), and T2DM status (A). Please refer to

Figure 1 legend for more details.


Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

standardized z-scores, or when applying different

meta-analysis methodology.

IL6 -174G�C and circulating glucose

The present analysis results of the IL6 -174G�Cwith fasting glucose was in line with our previous

joint analysis on T2DM, which had shown that

individuals carrying the IL6 -174 C-allele had 9%

lower odds for T2DM compared to individuals with

the GG-genotype (P�0.037) (25). We also con-firmed the dominant genetic model. Our previous

T2DM analysis had included a larger number of

subjects; when restricting to the seven studies with

fasting glucose data, a T2DM summary odds ratio

(OR) of 0.89 (95% CI�0.78�1.02) with a non-significant P-value of 0.09 would have been yielded,

which impressively demonstrates the gain from

additional information by using glucose as a quanti-

tative variable.

The recent finding of an association between the

IL6 -174 C-allele and T2DM (OR C-allele domi-

nant model�0.75 (95% CI�0.57�0.98)) in theFrench D.E.S.I.R. study including 307 T2DM cases

and 2919 normoglycemic controls (29) adds

strength to our hypothesis that IL6 -174G�C trulyhas an impact on individual T2DM risk.

IL6 -174G�C and body mass index

Several lines of evidence suggest that the cytokine

IL-6 plays a role in the regulation of body composi-

tion, probably by acting in a catabolic manner (30�32). To date, association between IL6 -174G�Cand obesity has been investigated by several com-

parably small studies with inconsistent results

(17,18,20,21,33). Qi et al. recently conducted a

large meta-analysis of mainly published studies

including 26,944 subjects and found statistically

significant heterogeneity between study-specific as-

sociation estimates. Applying a random effects

model in consequence, there was no evidence for

an association between IL6 -174G�C and BMI(34). Our joint analysis of 24,117 subjects also did

not find an association between IL6 -174G�C andBMI and added to the previous meta-analysis by sole

inclusion of individual-level data which enabled

application of completely standardized analysis

methods resulting in low between-study heterogene-

ity and by inclusion of mostly unpublished studies

(56% of the analyzed study participants). There was

only a small study overlap of 6631 subjects (27%)

between the present joint analysis and the Qi et al.

meta-analysis. Our power to detect a BMI difference

of 0.2 kg/m2 or 0.3 kg/m2 between IL6 -174 C-allele

carriers and non-carriers was more than 80% or

nearly 100%, respectively, given the type 1 error

probability of 0.050/3�0.017, a GG-genotype fre-quency of 31.2%, and a BMI standard deviation of

4.2 kg/m2.

IL6 -174G�C and circulating interleukin-6 levels

The current literature on the association between the

IL6 -174G�C and circulating IL-6 levels is incon-clusive. Several studies, with the largest including

641�1526 Caucasians (23,24,35�37), did not findevidence for association. However, other studies

with the largest being the recently published Cardi-

ovascular Health Study (CHS) with 4714 elderly

Caucasians (72 years median age) and a high

prevalence of T2DM (28%) showed borderline

significantly higher IL-6 levels for CC-genotype

subjects compared to GG (P�0.04) (38). In ourjoint analysis of 5659 Caucasians there was no

evidence for association between IL6 -174G�Cand circulating IL-6 levels including non-diabetic

and manifest T2DM subjects. Reasons for our non-

finding of an effect as compared to the CHS study

might be a smaller power in a joint analysis

compared to a single study, an effect only present

in special risk groups, or a true non-existence of the

effect. Our exploratory analysis of increased IL-6

levels with the CC-genotype in the 641 incident

T2DM subjects would be in line with the CHS

study. In summary, our joint analysis was not able to

resolve the puzzle whether and how IL-6 levels

mediate the association between IL6 -174G�Cand T2DM by pin-pointing a direct association

between the polymorphism and IL-6 levels. Thus,

even larger studies optimally with a prospective

design might be warranted.

Strengths and limitations of this joint analysis

The large number of subjects analyzed in our

investigation was a definite strength. For all analyzed

phenotypes, our study represents the largest study

on individual-level data conducted to date to address

the role of IL6 -174G�C. Thus, we were able toreveal small associations, which nevertheless are of

importance due to the high prevalence of the genetic

variant in the general population (41%�55%) andtheir potential to help understand the pathogenesis

of this severe disease. Our non-finding of an

association with BMI in this highly powered study

may also be of great importance to solve to some

extent the puzzle from contrary reports.

136 C. Huth et al.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Furthermore, it was a distinctive strength of our

joint analysis that it was based on individual

participants’ data allowing for standardized data

cleaning and analysis, to which the low heterogeneity

among study-specific estimates may be attributed.

Finally, our analysis included many unpublished

studies, which guards against the greatest threats of

meta-analyses, the publication or selective reporting

bias. Stratifying for published or unpublished studies

depicted a potential of some bias for the published

association studies of IL6 -174G�C and BMI.It may be considered a limitation that all T2DM

status-specific analyses exhibited truncated glucose

distributions, which consequently also affected the

correlated BMI and IL-6 level distributions. In

addition, no data were available on antidiabetic

medication for T2DM subjects for the glucose

analyses or on physical activity before blood extrac-

tion for the circulating IL-6 level analyses. This

might have decreased the precision of the estimates

but is unlikely to have caused false-positive findings.

Conclusion

This joint analysis represents by far the largest

individual participants’ data analysis to date on the

association of the IL6 promoter polymorphism

-174G�C with quantitative phenotypes relevantfor T2DM. Our data indicate that C-allele carriers

of the widely debated IL6 -174G�C polymorphismhave lower fasting glucose levels on average, which

substantiates previous findings of decreased T2DM

risk of these subjects. Therefore, our study suggests

IL6 as a T2DM gene. No statistically significant

association was found for quantitative BMI or IL-6

concentrations. In general, we demonstrated that a

consortium-based approach involving individual

participants’ data and standardized analysis is well

suited to investigate genetic variants with small

effects.

Acknowledgements

Funding: British Heart Foundation; Diabetes UK;

European Foundation for the Study of Diabetes;

German Diabetes Center; German Federal Ministry

of Education, Science, Research and Technology/

National Genome Research Network; German

Federal Ministry of Health and Social Security;

German Research Foundation; Helmholtz Zentrum

München; Ministry of Science and Research of the

State North Rhine-Westphalia; Munich Center of

Health Sciences; National Institutes of Health,

Swedish Research Council (no. K2007-54X-

09894-16-3), EC FP6 funding (contract no.

LSHM-CT-2003-503041), Sahlgrenska Center for

Cardiovascular and Metabolic Research.

We gratefully acknowledge the participation in the

original studies of all individuals used in this joint

analysis.

Declaration of interest: The authors report no

conflicts of interest. The authors alone are

responsible for the content and writing of the paper.

References

1. Wild S, Roglic G, Green A, Sicree R, King H. Global

prevalence of diabetes: estimates for the year 2000 and

projections for 2030. Diabetes Care. 2004;27:1047�53.2. Kempf K, Rathmann W, Herder C. Impaired glucose

regulation and type 2 diabetes in children and adolescents.

Diabetes Metab Res Rev. 2008 Jun 13 (Epub ahead of print).

3. Das SK, Elbein SC. The genetic basis of type 2 diabetes.

Cellscience. 2006;2:100�31.4. Carey AL, Bruce CR, Sacchetti M, Anderson MJ, Olsen DB,

Saltin B, et al. Interleukin-6 and tumor necrosis factor-alpha

are not increased in patients with Type 2 diabetes: evidence

that plasma interleukin-6 is related to fat mass and not insulin

responsiveness. Diabetologia. 2004;47:1029�37.5. Andreozzi F, Laratta E, Cardellini M, Marini MA, Lauro R,

Hribal ML, et al. Plasma interleukin-6 levels are indepen-

dently associated with insulin secretion in a cohort of Italian-

Caucasian nondiabetic subjects. Diabetes. 2006;55:2021�4.6. Henningsson S, Hakansson A, Westberg L, Baghaei F,

Rosmond R, Holm G, et al. Interleukin-6 gene polymorphism

-174G/C influences plasma lipid levels in women. Obesity.

2006;14:1868�73.7. Thorand B, Baumert J, Kolb H, Meisinger C, Chambless L,

Koenig W, et al. Sex Differences in the prediction of type 2

diabetes by inflammatory markers: Results from the MON-

ICA/KORA Augsburg case-cohort study, 1984�2002. Dia-betes Care. 2007;30:854�60.

8. Fishman D, Faulds G, Jeffery R, Mohamed-Ali V, Yudkin JS,

Humphries S, et al. The effect of novel polymorphisms in the

interleukin-6 (IL-6) gene on IL-6 transcription and plasma

IL-6 levels, and an association with systemic-onset juvenile

chronic arthritis. J Clin Invest. 1998;102:1369�76.9. Terry CF, Loukaci V, Green FR. Cooperative influence of

genetic polymorphisms on interleukin 6 transcriptional reg-

ulation. J Biol Chem. 2000;275:18138�44.10. Brull DJ, Montgomery HE, Sanders J, Dhamrait S, Luong L,

Rumley A, et al. Interleukin-6 gene -174g�c and -572g�cpromoter polymorphisms are strong predictors of plasma

interleukin-6 levels after coronary artery bypass surgery.

Arterioscler Thromb Vasc Biol. 2001;21:1458�63.11. Bennermo M, Held C, Stemme S, Ericsson CG, Silveira A,

Green F, et al. Genetic predisposition of the interleukin-6

response to inflammation: implications for a variety of major

diseases? Clin Chem. 2004;50:2136�40.12. Acalovschi D, Wiest T, Hartmann M, Farahmi M, Man-

smann U, Auffarth GU, et al. Multiple levels of regulation of

the interleukin-6 system in stroke. Stroke. 2003;34:1864�9.13. Gaudino M, Andreotti F, Zamparelli R, Di Castelnuovo A,

Nasso G, Burzotta F, et al. The -174G/C interleukin-6

polymorphism influences postoperative interleukin-6 levels

and postoperative atrial fibrillation. Is atrial fibrillation an

inflammatory complication? Circulation. 2003;108 Suppl

1:II195�9.


Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

14. Kilpinen S, Hulkkonen J, Wang XY, Hurme M. The

promoter polymorphism of the interleukin-6 gene regulates

interleukin-6 production in neonates but not in adults. Eur

Cytokine Netw. 2001;12:62�8.15. Vozarova B, Fernandez-Real JM, Knowler WC, Gallart L,

Hanson RL, Gruber JD, et al. The interleukin-6 (�174) G/Cpromoter polymorphism is associated with type-2 diabetes

mellitus in Native Americans and Caucasians. Hum Genet.

2003;112:409�13.16. Fernandez-Real JM, Broch M, Vendrell J, Gutierrez C,

Casamitjana R, Pugeat M, et al. Interleukin-6 gene poly-

morphism and insulin sensitivity. Diabetes. 2000;49:517�20.17. Berthier MT, Paradis AM, Tchernof A, Bergeron J, Prud’-

homme D, Despres JP, et al. The interleukin 6-174G/C

polymorphism is associated with indices of obesity in men. J

Hum Genet. 2003;48:14�19.18. Klipstein-Grobusch K, Möhlig M, Spranger J, Hoffmann K,

Rodrigues FU, Sharma AM, et al. Interleukin-6 g.-174G�Cpromoter polymorphism is associated with obesity in the

EPIC-Potsdam Study. Obes Res. 2006;14:14�18.19. Walch K, Grimm C, Zeillinger R, Huber JC, Nagele F, Hefler

LA. A common interleukin-6 gene promoter polymorphism

influences the clinical characteristics of women with poly-

cystic ovary syndrome. Fertil Steril. 2004;81:1638�41.20. Grallert H, Huth C, Kolz M, Meisinger C, Herder C,

Strassburger K, et al. IL-6 promoter polymorphisms and

quantitative traits related to the metabolic syndrome in

KORA S4. Exp Gerontol. 2006;41:737�45.21. Georges JL, Loukaci V, Poirier O, Evans A, Luc G, Arveiler

D, et al. Interleukin-6 gene polymorphisms and susceptibility

to myocardial infarction: the ECTIM study. Etude Cas-

Temoin de l’Infarctus du Myocarde. J Mol Med.

2001;79:300�5.22. Hamid YH, Rose CS, Urhammer SA, Glumer C, Nolsoe R,

Kristiansen OP, et al. Variations of the interleukin-6 promoter

are associated with features of the metabolic syndrome in

Caucasian Danes. Diabetologia. 2005;48:251�60.23. Sie MP, Sayed�Tabatabaei FA, Oei HH, Uitterlinden AG,

Pols HA, Hofman A, et al. Interleukin 6 -174 g/c promoter

polymorphism and risk of coronary heart disease: results

from the Rotterdam study and a meta-analysis. Arterioscler

Thromb Vasc Biol. 2006;26:212�7.24. Qi L, van Dam RM, Meigs JB, Manson JE, Hunter D, Hu

FB. Genetic variation in IL6 gene and type 2 diabetes:

tagging-SNP haplotype analysis in large-scale case-control

study and meta-analysis. Hum Mol Genet. 2006;15:1914�20.25. Huth C, Heid IM, Vollmert C, Gieger C, Grallert H, Wolford

JK, et al. IL6 gene promoter polymorphisms and type 2

diabetes: joint analysis of individual participants’ data from

21 studies. Diabetes. 2006;55:2915�21.26. Higgins JP, Thompson SG. Quantifying heterogeneity in a

meta-analysis. Stat Med. 2002;21:1539�58.27. Normand SL. Meta-analysis: formulating, evaluating, com-

bining, and reporting. Stat Med. 1999;18:321�59.28. Thakkinstian A, McElduff P, D’Este C, Duffy D, Attia J. A

method for meta-analysis of molecular association studies.

Stat Med. 2004;24:1291�306.29. Vaxillaire M, Veslot J, Dina C, Proenca C, Cauchi S,

Charpentier G, et al. Impact of common type 2 diabetes

risk polymorphisms in the DESIR prospective study. Dia-

betes. 2008;57:244�54.30. Wallenius V, Wallenius K, Ahren B, Rudling M, Carlsten H,

Dickson SL, et al. Interleukin-6-deficient mice develop

mature-onset obesity. Nat Med. 2002;8:75�9.31. van Hall G, Steensberg A, Sacchetti M, Fischer C, Keller C,

Schjerling P, et al. Interleukin-6 stimulates lipolysis and fat

oxidation in humans. J Clin Endocrinol Metab. 2003;88:

3005�10.32. Lyngso D, Simonsen L, Bulow J. Metabolic effects of

interleukin-6 in human splanchnic and adipose tissue. J

Physiol. 2002;543:379�86.33. Barbieri M, Rizzo MR, Papa M, Acampora R, De Angelis L,

Olivieri F, et al. Role of interaction between variants in the

PPARG and interleukin-6 genes on obesity related metabolic

risk factors. Exp Gerontol. 2005;40:599�604.34. Qi L, Zhang C, van Dam RM, Hu FB. Interleukin-6 genetic

variability and adiposity: associations in two prospective

cohorts and systematic review in 26,944 individuals. J Clin

Endocrinol Metab. 2007;92:3618�25.35. Bennet AM, Prince JA, Fei GZ, Lyrenas L, Huang Y, Wiman

B, et al. Interleukin-6 serum levels and genotypes influence

the risk for myocardial infarction. Atherosclerosis. 2003;171:

359�67.36. Herbert A, Liu C, Karamohamed S, Liu J, Manning A, Fox

CS, et al. BMI modifies associations of IL-6 genotypes with

insulin resistance: the Framingham Study. Obesity. 2006;14:

1454�61.37. van Oijen M, Arp PP, de Jong FJ, Hofman A, Koudstaal PJ,

Uitterlinden AG, et al. Polymorphisms in the interleukin 6

and transforming growth factor beta1 gene and risk of

dementia. The Rotterdam Study. Neurosci Lett. 2006;402:

113�7.38. Walston JD, Fallin MD, Cushman M, Lange L, Psaty B,

Jenny N, et al. IL-6 gene variation is associated with IL-6 and

C-reactive protein levels but not cardiovascular outcomes in

the Cardiovascular Health Study. Hum Genet. 2007;122:4

85�94.

138 C. Huth et al.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Appendix

Research design and methods

Study inclusion criteria, search strategy, and study

recruitment

All available published and unpublished studiesfulfilling the following criteria were recruited forthe IL6-T2DM consortium: 1) association studyconducted in humans, 2) polymorphic genotypedata for IL6 -174G�C, 3) T2DM cases and non-diabetic controls, 4) published before September2005 or unpublished, 5) availability of individualparticipants’ data (IPD). Studies were excluded ifthe control group consisted only of individuals withpre-diabetes (one study) (1), or if ethnic admixtureof unrelated study subjects was reported in theoriginal publication (Pima Indian case-control study,reported in (2)).

Published studies were identified in the PubMeddatabase using the following search terms: (IL-6 ORIL6 OR interleukin-6) AND (diabetes OR T2DMOR NIDDM) AND (gene OR genes OR genet* ORpolymorphism* OR allele*). To further extend thesearch, the reference lists from all identified originalstudies and review articles on this topic wereexamined. Unpublished studies were recruited by acall for participation at the symposium ‘Immunoge-netic Contribution to Type 2 Diabetes and Para-meters of the Metabolic Syndrome’, which was heldin September 2004 at the 40th Annual Meeting ofthe European Association for the Study of Diabetes,and by personally contacting investigators in thefield.

Data cleaning, phenotyping, and genotyping methods

The study center at the Helmholtz ZentrumMünchen checked all incoming data for plausibilityand for consistency with information provided by theinvestigators or the published article. Plausible andcorrected data were converted into a standardformat and incorporated into a central database.

Body mass index was calculated as weight (kg)divided by squared height (m2) from measuredanthropometric data, except for the RMIFAM andDANISH studies where self-reported data wereused. An overview on methods used to quantifycirculating interleukin-6 (IL-6) levels is presented inTable A-I.

A questionnaire was sent to all principal investi-gators to collect data on genotyping methods andquality. This information is also summarized inTable A-I. As IL6 -174G�C is a G/C-polymorph-ism, and allele G is complementary to allele C, thegenotyping sequences and strands, assessed via thequestionnaire, were compared with a reference toconfirm that the allele labeling was performedconsistently across all studies.

Further details on the statistical analyses

Statistical analyses were performed using SAS soft-ware version 9.1 (Cary, NC, USA). IL6 -174G�Callele and genotype frequencies were estimated,accounting for the correlation in family data by useof an exchangeable structure in a generalized esti-mating equations approach (SAS PROC GEN-MOD). Hardy-Weinberg equilibrium (HWE) wastested in non-diabetic subjects (SAS PROC AL-LELE); for family studies only one randomly drawnsubject per family was included. In order to approx-imate a normal distribution, circulating IL-6 waslogarithmically transformed.

Publication bias was investigated by visual in-spection of funnel plots and by the Egger’s regres-sion test (3). Funnel and forest plots were preparedusing Review Manager software version 4.2 (Co-chrane Collaboration, Copenhagen, DK).

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Figure A-1. (online appendix). Funnel plot for the association between IL6 -174G�C and body mass index (BMI). For each study, the ß-coefficient of the dominant model for the C-allele, adjusted for age, sex, and type 2 diabetes status, is plotted against its standard error as a

measure of study precision. All studies with IL6 -174G�C genotypes of non-diabetic individuals in Hardy-Weinberg equilibrium (HWE)are included. Black circles represent BMI published studies, white circles represent BMI unpublished studies. The vertical line marks the

pooled ß-coefficient (0.09 kg/m2).

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A-I. (online appendix). Methods used for quantifying circulating interleukin (IL)-6 levels and genotyping IL6 -174G�C (rs1800795according to http://www.ncbi.nlm.nih.gov).

Study IL-6 level quantification Genotyping method Call rate (%)a

Botnia Study n.a. Allelic discrimination

assay-by-design on ABI 7900

(Applied Biosystems)

100

Captopril Prevention Project Sandwich ELISA, R&D Systems,

Abingdon, UK

Dynamic allele specific

hybridization (DASH)

100

Danish Study n.a. Chip-based MALDI-TOF MS

(MassArray, Sequenom)

96

Ealing Diabetes Study of

Coagulation

n.a. Nla III RFLP, MADGE 97

European Prospective Investigation

into Cancer and Nutrition Potsdam

Sandwich ELISA, R&D Systems,

Abingdon, UK

SNuPE, MegaBACE 1000 100

The Finland-United States

Investigation of NIDDM Genetics

n.a. Illumina GoldenGate 98

Girona Genetics of Diabetes Study Solid-phase, enzyme-labeled,

chemiluminescent sequential

immunometric assay, DPC

DIPESA S.A., Madrid, Spain

SfaNI RFLP 99

KORA MI Family Study Sandwich ELISA, R&D Systems,

Abingdon,

UK

Hsp92II RFLP, PAGE 97

KORA Survey S4 Sandwich ELISA, CLB, Amsterdam,

Netherlands

Chip-based MALDI-TOF MS


97

KORA T2DM Family Study n.a. Chip-based MALDI-TOF MS


98

MONICA/KORA Case Cohort Study

S123

Sandwich ELISA, CLB,

Amsterdam,

Netherlands

Chip-based MALDI-TOF MS


99

MONICA/KORA Survey S3 n.a. Chip-based MALDI-TOF MS


99

Regensburg MI Family Study n.a. Hsp92II RFLP, PAGE 97

Second Northwick Park Heart Study n.a. PCR by MADGE, NIaIII RFLP 99

Tarraco Study n.a. SfaNI RFLP 99

University College Diabetes and

Cardiovascular Study

Sandwich ELISA, R&D Systems,

Abingdon, UK

PCR by MADGE, NIaIII RFLP 99

aSuccessfully genotyped individuals in percent of all subjects intended for genotyping.

n.a.�not applicable; NIDDM�non-insulin-dependent diabetes mellitus; MI�myocardial infarction.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A-II. (online appendix). Description of included studies.

Study name (official abbreviation) Study descriptiona

Botnia Study The Botnia Study began in 1990 as a family-based study aiming to identify genes increasing susceptibility to T2DM. Type 2

diabetic subjects from the area of five health care centers in the Botnia region of Western Finland were invited to participate

together with their family members. For the purpose of this joint analysis, unrelated individuals were genotyped for IL6 -174G�C. According to a priori defined criteria, one type 2 diabetic individual per family was selected. Non-diabetic subjects comprise

cases’ spouses and unrelated individuals, all being 35 years or older.

Captopril Prevention Project (CAPPP) CAPPP is a prospective randomized clinical trial conducted in Sweden and Finland during the 1990s. Patients aged 25�66 years,with a measured diastolic blood pressure of 100 mmHg or more on two occasions, were recruited at health centers and randomly

assigned to captopril or conventional antihypertensive treatment. Exclusion criteria were secondary hypertension, serum

creatinine concentration of more than 150 mmol/L, and disorders that required treatment with ß-blockers. Cases had T2DM atbase-line or were diagnosed during the follow-up. This joint analysis includes a substudy of the Swedish part of CAPPP, which has

been genotyped for IL6 -174G�C. This substudy comprises all patients that got myocardial infarction (MI), plus two controlsubjects without MI per patient, matched with respect to gender, age, and smoking. Further details: (4).

Danish Study The DANISH case-control study of T2DM involves all 4568 subjects with normal glucose tolerance (NGT) from the Inter99

cohort and 1389 unrelated type 2 diabetic patients recruited from the outpatient clinic at Steno Diabetes Center, Copenhagen and

the Research Center for Prevention and Health through the Inter99 study. The Inter99 cohort is a population-based randomized

non-pharmacological intervention study for prevention of cardiovascular disease done at the Research Center for Prevention and

Health involving 6514 Caucasian subjects (6164 with data from an oral glucose tolerance test). Further details: (5).

Ealing Diabetes Study of Coagulation (EDSC) The type 2 diabetic individuals of the EDSC study were recruited consecutively from the Ealing Hospital diabetes clinic in

London, UK. Patients completed a questionnaire with details of age, ethnicity, smoking habit, fasting status, duration of diabetes,

and other clinical details. Blood was collected for plasma and DNA analysis. Several further parameters, such as BMI, were

measured. Type 2 diabetic individuals (n�927) comprised primarily three ethnic groups: Indian Asian, n�503; UK white, n�331; black Afro-Caribbean, n�93. Further details: (6). To ensure comparability with the other Caucasian studies, only the whitesubjects were included in this joint analysis.

European Prospective Investigation into Cancer and Nutrition

Potsdam (EPIC-Potsdam)

A nested case-control study was designed within the European Prospective Investigation into Cancer and Nutrition Potsdam

cohort (EPIC-POTSDAM_nCC-T2DM), which is part of the European multicenter, population-based EPIC-study including

27,548 subjects from the area around Potsdam, Germany (women aged 35�65 years and men aged 40�65 years). Base-lineexamination and blood sampling were conducted between 1994 and 1998. Data presented in this joint analysis are based on the

first follow-up questionnaires sent to the study participants on average 2.3 years after base-line examination. Further details: (7).

To ensure comparability with the other cross-sectional studies, only the non-diabetic control subjects were included in this joint

analysis. Cases were free of T2DM at base-line and developed their incident T2DM during the follow-up. Analyses of their data

are presented separately.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A.II (Continued)


The Finland-United States Investigation of NIDDM Genetics

(FUSION)

The index probands in the FUSION study were identified primarily from the National Hospital Discharge Registry (NHDR),

which includes records since 1970 of all hospitalized patients with diabetes, and from previous studies carried out by the National

Public Health Institute in Finland. From the NHDR, all patients who were hospitalized with a diagnosis of T2DM in Finland

during 1987�1993 were identified in the first wave of sampling (FUSION 1). In the second wave of sampling (FUSION 2),patients hospitalized with T2DM during 1994�1995 were identified. Potential families for FUSION 2 also included someidentified during FUSION 1 but not invited to participate at that time due to distance from the study clinics. An index proband

with his family was eligible for participation in the FUSION study if 1) the proband or another affected sibling was diagnosed with

T2DM between 35 and 60 years of age, 2) there was no history of type 1 diabetes in first-degree relatives, 3) the proband had one

or more living full siblings diagnosed with T2DM at any age, and 4) at least one parent was apparently non-diabetic, with

preference given to families with living parents or parents who had lived a long life without known diabetes. Further details on

FUSION 1: (8); on FUSION 2: (9). Participants of FUSION 1 and FUSION 2 were analyzed separately. ‘FUSION 1’ in this joint

analysis comprises one type 2 diabetic individual from each FUSION 1 family, and non-diabetic spouses of type 2 diabetic

FUSION participants, as well as elderly subjects that were all born in 1925 and were normal glucose tolerant by oral glucose

tolerance tests (OGTTs) at both ages 65 and 70. ‘FUSION 2’ comprises the sibling generation of the FUSION 2 sampling wave.

Girona Genetics of Diabetes Study The type 2 diabetic patients of the Girona Genetics of Diabetes Study were consecutively recruited subjects from the diabetes

clinics at the Hospital of Girona, Spain. The non-diabetic subjects are unrelated healthy Caucasian middle-aged subjects recruited

from the general population. Further details: (2).

KORA Studies in chronological order KORA (Cooperative Health Research in the Region Augsburg) is a regional research platform in the German city of Augsburg and

the two adjacent counties, for population-based studies, subsequent follow-up studies, and family studies in the fields of

epidemiology, health economics, and health care research. KORA was established in 1996 to expand the WHO (World Health

Organization) MONICA (Monitoring of Trends and Determinants in Cardiovascular Disease) project in Augsburg. In the

framework of MONICA, three independent cross-sectional population-representative surveys were conducted in 1984/85 (S1),

1989/90 (S2), and 1994/95 (S3), and a population-based acute myocardial infarction registry was set up. The study subjects of all

Augsburg MONICA and KORA surveys and the family studies on myocardial infarction and T2DM are of German nationality

and were studied by physical examination, blood testing, and a standardized interview in KORA study centers. All tests were

carried out by specially trained personnel. Further details: (10�12). Some individuals were originally recruited for two or morestudies, but were assigned to one of the included KORA studies for the purpose of this joint analysis according to a priori defined

criteria.

MONICA/KORA Survey S3 The MONICA/KORA Survey S3 originally investigated 4856 individuals. Study participants that are included in the MONICA/

KORA Case Cohort Study S123 were eliminated from subjects of the MONICA/KORA Survey S3 for this joint analysis.

KORA MI Family Study Patients with MI prior to the age of 60 years and their siblings were identified through the acute myocardial infarction registry. The

diagnosis of MI was established according to the MONICA diagnostic criteria. Of 1254 patients contacted, 609 agreed to

participate in the study (532 men, aged 56.190.3 years). Moreover, 540 siblings without MI (251 men, aged 54.690.4 years from325 families) were recruited and examined by the same protocol.

KORA Survey S4 (KORA S4) The KORA S4 studied a population-representative sample of 4261 subjects, 25�74 years old, during the years 1999�2001. Thesample design followed the guidelines of the three previous MONICA Augsburg surveys. In the age-range of 55�74 years, 1653persons participated in an OGTT. These participants were genotyped for IL6 -174G�C and included in this joint analysis.Further details: (13).

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A.II (Continued)


KORA T2DM Family Study (KORA T2DMFAM) In 2001/2002, 605 nuclear families were enrolled in the KORA T2DM Family Study. Families were ascertained through an index

proband with known T2DM, who had at least one full sib or both parents willing to participate in the study. All available members

of the index probands’ nuclear families, i.e. full sibs and parents, were included. Index probands were all from the city or region of

Augsburg. They were recruited from T2DM patients of the Central Hospital of Augsburg, from earlier MONICA and KORA

studies, from the acute myocardial infarction registry, or via public relations. All participants were living in Germany, and all were

of European origin. Most subjects were extensively phenotyped in the KORA study center, some were examined by their family

doctor, who decided whether or not the subject had T2DM and took blood samples for DNA analyses. Data of the sibling

generation was included in this joint analysis.

MONICA/KORA Case Cohort Study S123 All participants of at least one of the three MONICA Augsburg surveys were prospectively followed for the MONICA/KORA Case

Cohort Study S123. The study was restricted to participants aged 35�74 years at base-line, since the incidence of T2DM is low inyounger subjects. A stratified random sample of the source population, containing 1885 subjects, was selected. A total of 555

incident cases of T2DM were observed between participants’ study start dates and 31st December 2002. Further details: (14). For

the purposes of this joint analysis on quantitative phenotypes, the base-line data of the MONICA/KORA Case Cohort Study S123

(prevalent T2DM and non-diabetic subjects) without the participants who developed their incident T2DM during the follow-up

were used to ensure comparability with the other cross-sectional studies (MONICA/KORA-BASE). Analyses of the subjects with

incident T2DM are presented separately.

Regensburg MI Family Study The kindreds of the Regensburg MI Family Study were ascertained through MI index patients, who were identified by screening

93,500 patient charts in seven cardiac in-hospital rehabilitation centers distributed throughout Germany. Index patients had all

suffered from MI before 60 years. If at least one sibling had suffered from MI or had severe coronary artery disease or by-pass

surgery, the index patient with all available parents and siblings were contacted and invited to participate in the study. All

participating individuals filled out a standardized questionnaire that focused on cardiovascular risk factors, medical diagnoses, life-

style and medication. Further details: (15). Data of the sibling generation were included in this joint analysis.

Second Northwick Park Heart Study (NPHS II) For the NPHS II Study, 3012 unrelated healthy Caucasian middle-aged male subjects were recruited from nine general medical

practices scattered throughout the UK and prospectively followed from 1989. Sixty-eight subjects with diabetes at base-line were

excluded from analysis. Further details: (16).

Tarraco Study For the Tarraco study, 211 unrelated type 2 diabetic subjects were recruited from the outpatient clinic at Hospital Universitari de

Tarragona ‘Joan XXIII’ during the years 2000�2004. Simultaneously, 118 healthy subjects were recruited from the same hospital.Further details: (2).

University College Diabetes and Cardiovascular Study

(UDACS)

The UDACS Study comprises 1011 consecutive subjects recruited from the diabetes clinic at University College London

Hospitals NHS Trust (UCLH) between the years 2001 and 2002. Patients completed a questionnaire with details of age, ethnicity,

smoking habit, fasting status, duration of diabetes, and other clinical details. Blood was collected for plasma and DNA analysis.

Several further parameters, such as BMI, were measured. No subjects requiring renal dialysis were recruited. Further details: (16).

aThe numbers of participants presented for the original studies do not always match the numbers used in the analyses of this joint analysis. The reason is that some subjects of the original studies

were not included in the joint analysis because of study overlap, missing genotype data, or missing phenotype data.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A-IV. (online appendix). Overview of studies and numbers of participants included in the analyses of the quantitative traits fasting

glucose, 2-h glucose, body mass index, and circulating interleukin (IL)-6 levels.

Fasting glucose 2-h glucose Body mass index Circulating IL-6 levels

Study Non-diaba T2DMa Fulla Non-diab T2DM Full Non-diab T2DM Full Non-diab T2DM Full

CAPPP 286 (37)b 323 424 (42)b 466 420 (42)b 462

DANISH 4397 1139 5536 4396 366 4762 4399 1212 5611

EDSC 106 299 299

EPIC-POTSDAM 65 348 348 346 346

FUSION 1 358 486 844 359 (37)b 396 367 508 875

FUSION 2 194 397 591 192 78 270 201 437 638

GIRONA 123 (41)b 164 89 (12)b 101 123 (42)b 165 84 (30)b 114

KORA-MIFAM 881 95 976 872 94 966

KORA-S4 1186 132 1318 1190 118 1308 1190 225 1415 1183 222 1405

KORA-T2DMFAM 512 152 664 505 56 561 513 776 1289

MONICA/KORA-BASE 1744 101 1845 1716 101 1817

MONICA-S3 299 3551 151 3702

NPHS II 2652 2652

RMIFAM 2614 662 3276

UDACS 560 560 549 549

Sum main analyses 7420 2412 9440 6731 618 7398 19007 5026 24117 4621 966 5659

Studies with Hardy-Weinberg equilibrium (HWE) violation (included in sensitivity analyses):

BOTNIA 557 728 1285 539 462 1001 557 731 1288

TGN 52 64 166 230

Sum all studies 7977 3192 10725 7270 1080 8399 19628 5923 25635 4621 966 5659

aNumber of ‘Non-diab’�non-diabetic subjects, ‘T2DM’�type 2 diabetic subjects, ‘Full’�non-diabetic and type 2 diabetic subjectscombined (used for full data analysis).bThese type 2 diabetic subjects were not included in the T2DM status-specific analysis because there were less than 50 type 2 diabetic

subjects in this study with data on the respective outcome.

Table A-III. (online appendix). References of included studies and overview on analyzed outcomes.

Study No. T2DM/Non-diaba Outcomeb Fasting glucose 2-h glucose BMI IL-6 Level Referencec

BOTNIA 731/557 3 3 3 4 n.a.

CAPPP 42/424 2 4 1 2 (4)

DANISH 1212/4399 1 3 1 4 (5)

EDSC 299/0 3 4 3 4 n.a.

EPIC-POTSDAM 0/348 4 4 2 2 (7)

FUSION 1 508/367 3 3 3 4 n.a.

FUSION 2 437/201 3 3 3 4 n.a.

GIRONA 42/123 3 3 3 3 (2)

KORA-MIFAM 95/881 4 4 2 2 (17)

KORA-S4 225/1190 3 3 1 1 (18)

KORA-T2DMFAM 776/513 3 3 3 4 n.a.

MONICA/KORA-BASE 101/1744 4 4 3 3 n.a.

MONICA-S3 151/3551 4 4 2 4 (17)

NPHS II 0/2652 4 4 1 4 (16)

RMIFAM 662/2614 4 4 2 4 (17)

TGN 166/64 4 4 3 4 (2)

UDACS 560/0 4 4 1 3 (16)

aNumber of type 2 diabetic (T2DM)/non-diabetic (non-diab) subjects included in analyses of the outcome BMI.bPublication of analyses on association between IL6 -174G�C and outcomes fasting glucose, 2-h glucose, body mass index (BMI), orinterleukin-6 (IL-6) levels 1�as main outcome, 2�as additional outcome; 3�completely unpublished before study recruitment for jointanalysis. 4�No or not enough outcome data available for analysis.cReferences of association studies between IL6 -174G�C and circulating glucose or IL-6 levels, BMI, or T2DM.n.a.�not applicable

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

Table A-V. (online appendix). Characteristics of study participants in the joint analysis.

Mean9standard error IL6 -174G�C

Study Statusa T2DM diagnosisb No.c Male (%) Age (years) BMI (kg/m2) % GC % CC P HWEd

BOTNIA 1 4, 10 731 53 60910 2995 53 25 0.182 2, 3, 4, 6 557 48 53912 2694 55 24 0.03

CAPPP 1 5, 7, 10 42 81 5895 3094 45 26 0.552 4 424 72 5797 2794 45 24 0.11

DANISH 1 3, 4, 5, 6, 10 1212 59 57911 3095 48 23 0.282 5, 7 4399 47 4598 2694 48 23 0.06

EDSC 1 3, 4, 10 299 63 63914 3096 45 22 0.132 n.a.

EPIC-POTSDAM 1 n.a.

2 1, 2 348 59 5597 2794 55 18 0.07FUSION 1 1 3, 4, 5, 6, 10 508 55 6397 3095 47 30 0.15

2 2, 3, 4, 7 367 41 6696 2794 50 30 1.00FUSION 2 1 3, 4, 5, 6, 10 437 55 6499 3095 49 29 0.89

2 2, 3, 4, 6 201 38 5899 2794 52 28 0.30GIRONA 1 5, 6, 10 42 76 53910 3195 50 10 0.73

2 4, 6 123 82 47913 2694 50 24 1.00KORA-MIFAM 1 1, 3, 8 95 85 5896 3095 50 15 0.66

2 1, 2, 8 881 67 5598 2894 47 17 0.71KORA-S4 1 2, 3, 5, 6, 10 225 59 6595 3195 54 16 0.17

2 1, 2, 4, 6 1190 51 6495 2894 52 17 0.06KORA-T2DMFAM 1 2, 3, 5, 6, 10 776 58 61910 3195 52 16 0.02

2 1, 2, 4, 6 513 43 58911 2895 52 16 0.17MONICA/KORA-BASE 1 2, 10 101 66 6297 3094 49 21 0.85

2 1 1744 53 52910 2794 49 17 0.55MONICA-S3 1 2, 3, 8, 10 151 58 6398 3094 46 23 0.33

2 1, 2, 8 3551 51 48914 2794 50 19 0.26NPHSII 1 n.a.

2 1 2652 100 5693 2693 50 18 0.47RMIFAM 1 1, 9 662 70 6197 2894 51 18 0.26

2 1, 9 2614 68 5899 2793 48 19 0.52TGN 1 1, 4 166 34 60910 2995 50 08 0.16

2 1, 3, 4, 6 64 41 49914 3095 61 08 0.04UDACS 1 4, 10 560 59 66911 3096 45 13 0.85

2 n.a.

a1�type 2 diabetic subjects, 2�non-diabetic subjects.bType of type 2 diabetes mellitus (T2DM)-diagnosis for type 2 diabetic subjects: 1�interview question; 2�interview question withdiabetes confirmation by doctor; 3�diabetes medication; 4�doctor diagnosis; 5�fasting glucose ]7.0 mmol/L; 6�oral glucosetolerance test with 2-hour glucose ]11.1 mmol/L (WHO-OGTT); 7�WHO-OGTT in some participants to confirm diagnosis; 8�random glucose ]11.1 mmol/L; 9�glycated hemoglobin (HbA1c) ]6.2%; 10�exclusion of subjects with type 1 diabetes.Type of ‘No T2DM’-diagnosis for non-diabetic subjects: 1�interview question; 2�no diabetes medication; 3�doctor diagnosis; 4�fasting glucoseB7.0 mmol/L; 5�fasting glucose B6.1 mmol/l; 6�oral glucose tolerance test (OGTT) with 2-hour glucoseB11.1 mmol/L; 7�OGTT with 2-hour glucose B7.8 mmol/L; 8�random glucose B11.1 mmol/L; 9�HbA1c B6.2%.cNumber of individuals included in analyses for association between IL6 -174G�C and body mass index (BMI).dP-value of exact test for Hardy-Weinberg equilibrium (HWE), using 10,000 Monte Carlo permutations. In the four family-studies

FUSION 2, KORA-MIFAM, KORA-T2DMFAM, and RMIFAM, HWE was calculated, including only one randomly drawn type 2

diabetic, or non-diabetic subject, respectively, per family.

n.a.�not applicable.

Ann

Med

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Uni

vers

ity o

f M

ichi

gan

Hea

lth S

yste

m

For

pers

onal

use

onl

y.

References

1. Kubaszek A, Pihlajamaki J, Komarovski V, Lindi V, Lindstrom

J, Eriksson J, et al. Promoter polymorphisms of the TNF-alpha

(G-308A) and IL-6 (C-174G) genes predict the conversion

from impaired glucose tolerance to type 2 diabetes: the Finnish

Diabetes Prevention Study. Diabetes. 2003;52:1872�6.2. Vozarova B, Fernandez-Real JM, Knowler WC, Gallart L,

Hanson RL, Gruber JD, et al. The interleukin-6 (�174) G/Cpromoter polymorphism is associated with type-2 diabetes

mellitus in Native Americans and Caucasians. Hum Genet.

2003;112:409�13.3. Egger M, Davey SG, Schneider M, Minder C. Bias in meta-

analysis detected by a simple, graphical test. BMJ. 1997;315:

629�34.4. Wernstedt I, Eriksson AL, Berndtsson A, Hoffstedt J, Skrtic S,

Hedner T, et al. A common polymorphism in the interleukin-6

gene promoter is associated with overweight. Int J Obes Relat

Metab Disord. 2004;28:1272�9.5. Hamid YH, Rose CS, Urhammer SA, Glumer C, Nolsoe R,

Kristiansen OP, et al. Variations of the interleukin-6 promoter

are associated with features of the metabolic syndrome in

Caucasian Danes. Diabetologia. 2005;48:251�60.6. Flavell DM, Ireland H, Stephens JW, Hawe E, Acharya J,

Mather H, et al. Peroxisome proliferator-activated receptor

alpha gene variation influences age of onset and progression of

type 2 diabetes. Diabetes. 2005;54:582�6.

7. Mohlig M, Boeing H, Spranger J, Osterhoff M, Kroke A,

Fisher E, et al. Body mass index and C-174G interleukin-6

promoter polymorphism interact in predicting type 2 diabetes.

J Clin Endocrinol Metab. 2004;89:1885�90.8. Valle T, Tuomilehto J, Bergman RN, Ghosh S, Hauser ER,

Eriksson J, et al. Mapping genes for NIDDM. Design of the

Finland-United States Investigation of NIDDM Genetics

(FUSION) Study. Diabetes Care. 1998;21:949�58.9. Silander K, Mohlke KL, Scott LJ, Peck EC, Hollstein P, Skol

AD, et al. Genetic variation near the hepatocyte nuclear

factor-4 alpha gene predicts susceptibility to type 2 diabetes.

Diabetes. 2004;53:1141�9.10. Wichmann HE, Gieger C, Illig T. KORA-gen*resource for

population genetics, controls and a broad spectrum of disease

phenotypes. Gesundheitswesen. 2005;67 Suppl 1:S26�30.11. Holle R, Happich M, Löwel H, Wichmann HE. KORA*a

research platform for population based health research.

Gesundheitswesen. 2005; 67 Suppl 1:S19�25.12. Lowel H, Doring A, Schneider A, Heier M, Thorand B,

Meisinger C. The MONICA Augsburg surveys*basis forprospective cohort studies. Gesundheitswesen. 2005;67

Suppl 1:S13�18.13. Rathmann W, Haastert B, Icks A, Löwel H, Meisinger C,

Holle R, et al. High prevalence of undiagnosed diabetes

mellitus in Southern Germany: Target populations for

efficient screening. The KORA survey 2000. Diabetologia.

2003;46:182�9.

Table A-VII. (online appendix). Full data sensitivity analyses, additionally including studies showing Hardy-Weinberg equilibrium (HWE)

violation in non-diabetic subjects (C-allele dominant model).

Outcome

Additionally included

studies Individual n

Individual b-estimates(95% CI) Su

Joint analysis of individual participants data from 17...

Documents

Transcript of Joint analysis of individual participants data from 17...